<%BANNER%>
Power to identify a genetic predictor of antihypertensive drug response using different methods to measure blood pressur...
CITATION DOWNLOADS PDF VIEWER
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/AA00012396/00001
 Material Information
Title: Power to identify a genetic predictor of antihypertensive drug response using different methods to measure blood pressure response
Series Title: Journal of Translational Medicine
Physical Description: Book
Language: English
Creator: Turner, Stephen T.
Schwartz, Gary L.
Chapman, Arlene B.
Beitelshees, Amber L.
Gums, John G.
Cooper-DeHoff, Rhonda M.
Boerwinkle, Eric
Johnson, Julie A.
Bailey, Kent R.
Publisher: BioMed Central
Publication Date: 2012
 Subjects
Subjects / Keywords: hypertension
blood pressure monitoring
antihypertensive drug therapy
beta-blocker
thiazide diuretic
plasma renin activity
 Notes
Abstract: Background: To determine whether office, home, ambulatory daytime and nighttime blood pressure (BP) responses to antihypertensive drug therapy measure the same signal and which method provides greatest power to identify genetic predictors of BP response. Methods: We analyzed office, home, ambulatory daytime and nighttime BP responses in hypertensive adults randomized to atenolol (N = 242) or hydrochlorothiazide (N = 257) in the Pharmacogenomic Evaluation of Antihypertensive Responses Study. Since different measured BP responses may have different predictors, we tested the “same signal” model by using linear regression methods to determine whether known predictors of BP response depend on the method of BP measurement. We estimated signal-to-noise ratios and compared power to identify a genetic polymorphism predicting BP response measured by each method separately and by weighted averages of multiple methods. Results: After adjustment for pretreatment BP level, known predictors of BP response including plasma renin activity, race, and sex were independent of the method of BP measurement. Signal-to-noise ratios were more than 2-fold greater for home and ambulatory daytime BP responses than for office and ambulatory nighttime BP responses and up to 11-fold greater for weighted averages of all four methods. Power to identify a genetic polymorphism predicting BP response was directly related to the signal-to-noise ratio and, therefore, greatest with the weighted averages. Conclusion: Since different methods of measuring BP response to antihypertensive drug therapy measure the same signal, weighted averages of the BP responses measured by multiple methods minimize measurement error and optimize power to identify genetic predictors of BP response.
 Record Information
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution.
Resource Identifier: doi - 10.1186/1479-5876-10-47
System ID: AA00012396:00001

Downloads

This item is only available as the following downloads:

Article ( PDF )

Table S1 ( DOC )

Eprints Dublin Core METS ( XML )

BioMed Central XML


Full Text

PAGE 1

RESEARCH OpenAccessPowertoidentifyageneticpredictorof antihypertensivedrugresponseusingdifferent methodstomeasurebloodpressureresponseStephenTTurner1*,GaryLSchwartz1,ArleneBChapman3,AmberLBeitelshees4,JohnGGums5,6, RhondaMCooper-DeHoff5,7,EricBoerwinkle8,JulieAJohnson5,7andKentRBailey2AbstractBackground: Todeterminewhetheroffice,home,ambulatorydaytimeandnighttimebloodpressure(BP) responsestoantihypertensivedrugtherapymeasurethesamesignalandwhichmethodprovidesgreatestpower toidentifygeneticpredictorsofBPresponse. Methods: Weanalyzedoffice,home,ambulatorydaytimeandnighttimeBPresponsesinhypertensiveadults randomizedtoatenolol(N=242)orhydrochlorothiazide(N=257)inthePharmacogenomicEvaluationof AntihypertensiveResponsesStudy.SincedifferentmeasuredBPresponsesmayhavedifferentpredictors,wetested the samesignal modelbyusinglinearregressionmethodstodeterminewhetherknownpredictorsofBP responsedependonthemethodofBPmeasurement.Weestimatedsignal-to-noiseratiosandcomparedpowerto identifyageneticpolymorphismpredictingBPresponsemeasuredbyeachmethodseparatelyandbyweighted averagesofmultiplemethods. Results: AfteradjustmentforpretreatmentBPlevel,knownpredictorsofBPresponseincludingplasmarenin activity,race,andsexwereindependentofthemethodofBPmeasurement.Signal-to-noiseratiosweremorethan 2-foldgreaterforhomeandambulatorydaytimeBPresponsesthanforofficeandambulatorynighttimeBP responsesandupto11-foldgreaterforweightedaveragesofallfourmethods.Powertoidentifyagenetic polymorphismpredictingBPresponsewasdirectlyrelatedtothesignal-to-noiseratioand,therefore,greatestwith theweightedaverages. Conclusion: SincedifferentmethodsofmeasuringBPresponsetoantihypertensivedrugtherapymeasurethe samesignal,weightedaveragesoftheBPresponsesmeasuredbymultiplemethodsminimizemeasurementerror andoptimizepowertoidentifygeneticpredictorsofBPresponse. Keywords: hypertension,bloodpressuremonitoring,antihypertensivedrugtherapy,beta-blocker,thiazidediuretic, plasmareninactivityBackgroundAlthoughofficebloodpressure(BP)measurements remainthestandard-of-care,averagesofout-of-office measurementsaremorereproducible[1].Out-of-office averageshavealsobeenreportedtobemorestrongly correlatedwithsubclinical targetorgandamage[2,3] andtobetterpredictfuturecardiovasculardisease events[4-6]thanofficemeasurements.Notsurprisingly, BPresponsestoantihypertensivedrugtherapyaremore preciselyandaccuratelyde terminedbyout-of-office thanofficemeasurements,whichareinfluencedby whitecoatandplaceboeffects[7,8].Consequently, greateruseofout-of-officemethodsofBPmeasurement hasbeenadvocatedforclinicaldecision-makingand research[1]. Moreindividualizedapproachestoantihypertensive drugtherapymaybecomepossibleifgeneticpolymorphismsarediscoveredthatimprovetheabilityto *Correspondence:sturner@mayo.edu1DivisionofNephrologyandHypertension,DepartmentofMedicine,Mayo Clinic,Rochester,MN55905,USA FulllistofauthorinformationisavailableattheendofthearticleTurner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 2012Turneretal;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsoftheCreativeCommons AttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,distribution,andreproductionin anymedium,providedtheoriginalworkisproperlycited.

PAGE 2

predictinter-individualdifferencesinBPresponse[9]. Knownpredictorsarelimitedtorace,age,andplasma reninactivity[10,11],whichexplainlessthan50%of interindividualvariationinBPresponsetosingle-drug therapy[12,13].Mostpreviousstudieshaveattempted toidentifygeneticornon-geneticpredictorsofofficeBP response,whichisnotveryreproducibleandcorrelates onlymodestlywithhomeandambulatoryBPresponses [7,8,14-16].Whetherout-of-officemeasurementsofBP responsecanimprovetheabilitytoidentifypredictors ofBPresponsehasnotbeendemonstrated.Method-specificmeasurementerrorscouldaccountfordifferences inthemagnitudeofandcorrelationbetweenoffice, home,ambulatorydaytimeandnighttimeBPresponses [8].However,anadditionalpo ssibilityisthatdifferent BPresponsesignalsaremeasuredbythedifferent methods. SincedifferentBPresponsesignalsmayhavedifferent predictors,ourfirstobjec tiveinthepresentstudywas totestthe samesignal modelbydeterminingwhether knownpredictorsofBPresponse,i.e.,race,age,and plasmareninactivity,dependonthemethodofBPmeasurement.WeanalyzeddatafromthePharmacogenomic EvaluationofAntihypertensiveResponses(PEAR)study, inwhichBPresponsestosingle-drugtherapywithatenololorhydrochlorothiazideweremeasuredbyallfour methods[7,8].Inthiscontext,oursecondobjectivewas toestimatesignal-to-noiseratiosandcomparethe powertoidentifyageneticpolymorphismpredictingBP responsewhenmeasuredbyeachmethodseparatelyand byweightedaveragesofmultiplemethods.MethodsParticipantsThePEARstudy[17]http://cli nicaltrials.gov/ct2/show/ NCT00246519wasapprovedbytheInstitutionalReview Boardateachsite,andallparticipantsgaveinformed consent.Ataninitialconsentandscreeningvisit, trainedstudypersonneladministeredstandardizedquestionnaires,performedalim itedphysicalexamination, andobtainedbloodandurinesamplesfortestingto establisheligibilityforparticipation[11].Participants wereprovidedanautomatedsphygmomanometer (MicroLife3AC1-PC,MinneapolisMN),theadequacy ofwhichhasbeenpreviouslyvalidated[18],andwithdrawnfrompreviousantihypertensivedrugtherapy.The devicewassettomeasureBPintriplicatewitheach activationandtostoretheaveragesystolicanddiastolic BPsandthetimeofeachsetofmeasurements.Participantswereinstructedtotakereadingsdailyinthe seatedposition,onesetofthreereadingsinthemorning uponarisingfrombedandasecondsetintheevening justbeforeretiring.Atsubsequentstudyvisits(priorto randomizationandattheendoftherapy),anadditional setofthreereadingswasobtainedseated(>5minutes) intheofficeusingthehomemonitor.Inaddition,24hourambulatoryBPrecordingswereobtainedatthese visitsusingSpacelabs(RedmondWA)ambulatorymonitors,model90207,theadequacyofwhichhasbeenpreviouslyvalidated[19].Participantswereinstructedto conducttheirusualdailyactivitieswhilewearingthe monitor,whichwassettorecordBPfourtimesper hourduringtheday(6AMto10PM)andtwiceper hourduringthenight(10PMto6AM).Theaverage( standarddeviation)numb erofambulatorymeasurementswas6710duringdaytimehoursand153 duringnighttimehours. Attheendofthedrug-freewashoutperiod,fasting bloodsamplesweredrawnintheseatedpositionafter ambulationformeasurementofplasmareninactivity [11].Toqualifyforrandom ization,theaveragehome diastolicBPinthepreviousweekhadtobe 85mmHg (consistingofatleastfivemorningandfiveeveningsets ofreadings) and theaverageofficediastolicBP 90 mmHg.Participantsreceivedeitheratenololorhydrochlorothiazide,startingat50mgor12.5mgdaily, respectively,fortwoweeks,afterwhich,ifBPremained >120/70mmHg,thedoseswereincreasedto100mgor 25mgdaily,respectively,forsixadditionalweeks.StatisticalanalysisAnalyseswereperformedwithStatisticalAnalysisSystemsoftware,version9.1(SAS,Raleigh-DurhamNC). Statisticalsignificancewasdefinedaprioriby P <0.05. TheBPresponsetoeachdrugwascalculatedforeach measurementmethodbysubtractingthepretreatment averagefromthepost-treatmentaverage.ThehomeBP averagesconsistedofatleastfiveofsevenmorningand eveningsetsofthreereadingstakenduringtheweek priortothepre-andpost-treatmentstudyvisits(i.e.,at least30andupto42measurementspriortoeachvisit). Multiple-variablelinearregressionanalyseswereperformedtoidentifyparticipantcharacteristicsthatmade additive,statisticallyindependentcontributionstothe predictionofsystolicanddiastolicBPresponsetoeach drug.Inpreliminaryanalyses,wefoundthathigherpretreatmentBPlevelwasassociatedwithgreaterBP response,asexpected[20].Becausewesoughttoevaluatepredictorsthatareindependentofthepretreatment BPlevel,wefirstregressedouttheeffectsofpretreatmentBPlevelandthenmodeledtheeffectsofother knownpredictorsofBPresponse[10,11]aswellas othervariablesmeasuredattheconsentandscreening visit[11,17].Finalmultiple-variablemodelswerederived usingabackwardstepwiseeliminationprocedure, retainingonlythepredictorsofbothsystolicanddiastolicBPresponsestoeitherdrug.Intheinitialmodels thatincludedraceandpretreatmentplasmareninTurner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page2of9

PAGE 3

activity,agewasnotastatisti callysignificantpredictor ofBPresponsesandwasnotretainedinthefinal models. TodeterminewhetherknownpredictorsofBP responsedependonthemethodofmeasuringBP response,wecomparedmodelsinwhichregression coefficientswereconstrainedtobeidenticalamong measurementmethodsorallowedtodifferamongmethodsbyincludinginteractionsofeachpredictorwiththe methodofBPmeasurement. Consideringthemodelin whichtheregressioncoefficientswereidenticalacross methodsasanullhypothesis,weattemptedtodetect anydeparturesindicatingdependencyofthepredictors onmethodofBPmeasurementthatwouldleadusto rejectthe samesignal model.ThisanalysisusedPROC GENMODinSAS,whichadj ustsforthecorrelation amongthefourBPresponsemeasurementswithineach participant. Toestimatesignal-to-noiseratios,thecovariance matrixofthefourmeasuredBPresponseswasusedto estimatethesignalandnoisecomponentsforeach methodofmeasuringBPresponseafterregressingout themethod-specificeffectsofpretreatmentBPlevel. ThecorrelationcoefficientbetweenBPresponsesmeasuredbytwomethodsprovidesadimensionlessmeasure ofhowmuchthetworesponsescovary(change together);thecovariancebetweenthemexpressesthe correlationinunitsofthetwoBPresponsesmultiplied together(mm2Hg)andisthevariancesharedbetween them,i.e.,thesignalvariance.Sinceeachpair-wisecovarianceprovidesanunbiasedestimateofthesignalvariance,weusedtheaverageofthesixpairwise covariancesastheBPresponsesignal.Subtractingthe signalfromthemethod-specifictotalvarianceprovided anestimateofthemethod-specificerrorvarianceor noise. Weexaminedimplicationsofthesignal-to-noiseanalysesforaccomplishingthegoalofthePEARandother pharmacogenomicstudies.S pecifically,wecompared powerandsamplessizesrequiredtoidentifyagenetic polymorphismthatpredictsBPresponsewhenmeasured byeachmethodseparatelyan dbyweightedaveragesof theresponsesmeasuredbymultiplemethods.Therationalefortheweightedaverageswastoincreasethesignalto-noiseratio(andpower)byminimizingtheerrorvariance.TwodifferentcombinationsofthemeasuredBP responseswereconsidered:aweightedaverageofallfour methodsandaweightedaverageoftheofficeandhome BPresponses.Theweighteda veragecombinationswere determinedbasedontherowsumsoftheinverseofthe inter-methodcovariancematrices,whichprovideweights thatminimizethevariance[21]. Forthepowerandsamplessizecalculations,we assumedthatageneticpolymorphismwithaminor allelefrequencyof0.2influencestheBPresponsesignal withaneffectsizethatcanbedetectedwith80%power inasampleofN=300atagenome-widesignificance levelof50-8.This P -valuewasoriginallysuggested forgenome-wideassociationanalysisof1millionsingle nucleotidepolymorphismsusingaBonferronicorrection formultipletesting[22].Basedontheestimatedsignal variances,wecalculatedthealleleeffectsizes(in mmHg/allele)andthepercentageofvariationintheBP response( R2100%)explainedbythepolymorphism. WethencalculatedthepowertodetectthepolymorphisminasampleofN=300whentheBPresponseis measuredbyeachmethodseparatelyandweighted averagesofmultiplemethods,andthecorresponding samplesizesrequiredtomaintain80%power.ResultsSampledescriptionFivehundredandninety-fiv estudyparticipantshad completemeasurementsofoffice,home,andambulatory daytimeandnighttimeBPresponses(Table1).Ofthese, 293participantswererandomizedtoatenolol(49%)and 302tohydrochlorothiazide(51%)treatment.Mean valuesandrelativefrequenciesofparticipantcharacteristicsmeasuredpriortorandomizationdidnotdiffersignificantlybetweentheatenololandhydrochlorothiazidetreatedgroups[11](notshown).Office,home,andambulatoryBPresponseMeansandstandarddeviationsofthesystolicanddiastolicBPresponsesdifferedamongmeasurementmethods(Table2).ForsystolicBPresponse,office measurementshadthegreatestmeandeclinesandhome measurementsthesmallestmeandeclinesinresponseto eachdrug(Table2).FordiastolicBPresponse,office measurementsalsohadthegreatestmeandeclinesin responsetoeachdrug;ambulatorynighttimemeasurementshadthesmallestmeandeclineinresponsetoatenololandhomemeasurementsthesmallestmean declineinresponsetohydrochlorothiazide.Correlation coefficientsbetweentheoffice,home,ambulatorydaytimeandnighttimeBPresponsesweremodestinmagnitude(notshown),rangingfrom0.36to0.71after adjustmentfordifferencesinpretreatmentBPlevels(all P <0.0001).TheBPresponsesignalanditspredictorsWeassessedwhethertheoffice,home,andambulatory daytimeandnighttimeBPresponsesmeasurethesame BPresponsesignalbydeterminingwhetherthepredictorsofBPresponsedependonthemethodofBPmeasurement(seeMethods).Afteradjustmentfor pretreatmentBPlevel,noneofthepredictorsofBP responsedependeduponofthemethodofBPTurner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page3of9

PAGE 4

measurement(analysesnotshown).Forallfourmethods ofmeasuringBPresponse,theidentifiedpredictors includedrace,plasmareninactivity,andsex(Table3). Asexpected,blackracewasassociatedwithlessersystolicanddiastolicBPresponse stoatenololandgreater responsestohydrochlorothiazide;andgreaterlogrenin wasassociatedwithgreatersystolicanddiastolicBP responsestoatenololandlesserresponsestohydrochlorothiazide[11].Males exwasindependentlyassociatedwithlessersystolicanddiastolicBPresponsesto eachdrug.Greaterloghypertensionyearsandgreater serumALTwereeachindependentlyassociatedwith greatersystolicanddiastolicBPresponsestoatenolol butnottohydrochlorothiazide.Signal-tonoise-ratiosInferringthatallfourmethodsmeasurethesameBP responsesignal,weestimatedthesignalvariance(see Methods)andcalculatedthemethod-specificerrorvariance(noise)andsignal-to-noiseratioforeachmeasured BPresponse(Figures1and2).Thehomeandambulatory daytimeBPresponseshadthelargestsignal-to-noiseratios andtheambulatorynighttimeandofficeBPresponsesthe smallestsignal-to-noiseratios(Figure2).Thesignal-tonoiseratiosofthehomeandambulatorydaytimeBP responsesweresimilarinmagnitudeandupto4-fold greaterthanthesignal-to-noiseratiosoftheofficeand ambulatorynighttimeBPres ponses,whichweremostly lessthan1(morenoisethansignal).Weightedaveragesof Table1DescriptivecharacteristicsofstudyparticipantsMeanstandarddeviationorN(%) N(%)595(100) Randomizedtohydrochlorothiazide,N(%) 302(51) Age,years 49.39.1 Male,N(%) 280(47) Black,N(%) 245(41) BMI,kgm-230.65.6 Hypertensionduration,years 7.17.2 Antihypertensivemedication,N(%) 492(91) Currentsmoker,N(%) 69(13) Glucose,mgdL-194.810.5 Creatinine,mgdL-10.90.2 SerumALT,UL-129.115.9 Plasmareninactivity,ngmL-1hr-11.01.2 Restingheartrate,beatmin-171.010.2 ScreeningofficesystolicBP,mmHg 137.913.8 ScreeningofficediastolicBP,mmHg 89.58.8 Pretreatmentsystolicbloodpressure,mmHg 151.513.8 Pretreatmentdiastolicbloodpressure,mmHg 98.26.3BMI,bodymassindex;ALT,alanineaminotransferase;BP,mmHg.Characteristicsweremeasuredatthescreeningvisitexceptpretreatmentsystolicanddiastolic BPandplasmareninactivityweremeasuredattheendofthedrug-freewashoutperiodpriortoinitiatingatenololorhydrochlorothiazidetherapy. Table2BloodPressureResponsestoMonotherapybyMeasurementMethodAllN=595 AtenololN=293 HydrochlorothiazideN=302 SystolicBPResponse,mmHg Office-13.414.7 -13.515.6 -13.213.7 Home-8.89.8 -8.310.4 -9.49.1 Ambulatorydaytime-11.510.5 -12.211.1 -10.89.8 Ambulatorynighttime-9.712.5 -8.912.9 -10.612.1 Contrast P value<0.001 <0.001 <0.001 DiastolicBPResponse,mmHg Office-8.68.9 -10.59.4 -6.87.9 Home-6.66.5 -7.86.8 -5.36.0 Ambulatorydaytime-7.67.6 -9.27.9 -6.17.0 Ambulatorynighttime-6.89.5 -7.010.0 -6.59.0 Contrast P value<0.001 <0.001 0.10BP,bloodpressure.Turner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page4of9

PAGE 5

allfourmeasuredBPresponsesimprovedthesignal-tonoiseratiosupto4-foldcomparedtothehomeBP responsesandupto19-foldcomparedtotheofficeBP responses.WeightedaveragesofthehomeandofficeBP responsesimprovedthesignal-to-noiseratiosmodestly comparedtothehomeBPresponses(by24%atmost). Theweightings,providedinAdditionalfile1:TableS1, minimizedtheerrorvariance(noise)oftheaverageBP responses(seeMethods),therebyaccountingforthe improvementinsignal-to-noiseratios.Powerandsamplesizeneededtoidentifyagenetic predictorofbloodpressureresponseWeassumedthatageneticpolymorphismwithminor allelefrequencyof0.2influencestheBPresponsesignal andisdetectedwith80%poweratgenome-widesignificancelevelof510-8(seeMethods)[22].TheT-statisticforassociationofthepolymorphismwithBP responseis6.29;11.6%ofthesignalvariationis explained( R2100%);andthecorrespondingeffectsizes ( b -coefficients)inmmHgperalleleare5.22/3.36for thesystolic/diastolicBPresponsestoatenololand4.04/ 2.63forthesystolic/diastolicBPresponsestohydrochlorothiazide.Basedonthesignalandnoiseanalyses, powertodetectthispolymorphisminasamplesizeof N=300declinedforallmethodsofmeasuringBP responsewhencomparedtoaperfectmethodcapable ofmeasuringonlysignalandnonoise(Table4).Power declinedmostmarkedlyfortheofficeBPresponses(to <5%power)andwasonlymaintainedat>50%forthe Table3Multi-variablelinearregressionmodelingofpredictorsofbloodpressureresponsesignalafteradjustmentfor pretreatmentbloodpressurelevelBPResponsetoAtenolol(N=293) BPResponsetoHydrochlorothiazide(N=302) Systolic b SEDiastolic b SE Systolic b SE Diastolic b SE Intercept -14.10.9 -11.80.6 -11.80.7 -6.70.5 Race:Black 6.31.1 4.80.8 -2.90.9 -2.00.6 LogPlasmaReninActivity -4.40.5 -2.50.3 0.70.5 0.70.3* Sex:Male 2.31.0* 2.90.7 4.20.8 2.70.5 LogHypertensionYears 1.30.5* 1.00.4 -SerumALT 0.080.03 0.040.02* Model R2%22% 23% 11% 16%BP,bloodpressure; b ,regressioncoefficient;SE,standarderror.Modelparametersareestimatedatthemeanvaluesforeachquantitativepredictorvariableina combineddatasetusedtomodelthepredictorsofoffice,home,ambulatorydaytimeandnighttimeBPresponsesafteradjustmentforpretreatmentBPlevels. P valuesfortestsofmodelparameters=0:*, 0.05; 0.01; 0.001;, 0.0001. R2%isthepercentageofvariationintheofficeBPresponseexplainedby themodelpredictors. Figure1 Signal,noise,andtotalvariancesofthemeasuredbloodpr essureresponsestosingle-drugtherapywithatenololor hydrochlorothiazideandtheirweightedaverages Turner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page5of9

PAGE 6

weightedaveragesofallfourmethods.Alternatively,to maintain80%powerthesamplesizeswouldneedtobe increasedforallmethodsofmeasuringBPresponse whencomparedtotheN=300samplesizefortheperfectmeasurementofsignalwithoutnoise(Table4).The samplesizesincreasedmostmarkedlyfortheofficeBP responses(by>200%),butonlymodestlyforthe weightedaveragesofallfourmethods(by 26%).DiscussionOurfirstobjectivewastoassesswhetheroffice,home, andambulatorydaytimeandnighttimemeasurementsof Figure2 Signal-to-noiseratiosofthemeasuredbloodpressureresponsestosingle-drugtherapywithatenololorhydrochlorothiazide andtheirweightedaverages Table4PowerandsamplesizestodetectsinglenucleotidepolymorphisminfluencingBPresponsemeasuredby office,home,ambulatorydaytimeandnighttimebloodpressureDrug AtenololHydrochlorothiazide Power,N=300N,80%powerPower,N=300N,80%power SystolicBPresponsesignal80%30080%300 Measurementmethods Office 4% 780 1.5% 1013 Home 42% 416 24% 500 Ambulatoryday 41% 420 18% 540 Ambulatorynight 17% 552 12% 607 Weightedaverages Allmethods 70% 328 61% 353 Homeandoffice 45% 404 30% 467 DiastolicBPresponsesignal 80% 300 80% 300 Measurementmethods Office 5% 764 3% 884 Home 39% 426 19% 537 Ambulatoryday 39% 426 19% 537 Ambulatorynight 27% 484 7% 695 Weightedaverages Allmethods 63% 348 379 Homeandoffice 42% 416 26% 489BP,bloodpressure.Thepowerandsamplesizeestimatesareforasinglenucleotidepolymorphismwithminorallelefrequencyof0.2thatinfluencesth eBP responsesignalwithaneffectsizedetectedwith80%powerinasampleofN=300atagenome-widesignificancelevelof510-8(seeMethods).Theestimates assumethattheBPresponsesignalcanbemeasuredwithouterror.Turner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page6of9

PAGE 7

BPresponsetosingle-dru gtherapymeasurethesame BPresponsesignal.SincedifferentBPresponsesignals mayhavedifferentpredictors,thisassessmentwasbased ondeterminingwhetherknownpredictorsofBP response[11]dependonthemethodofBPmeasurement.Afteradjustmentforthemethod-specificeffects ofpretreatmentBPlevel,theidentifiedpredictorsofBP responsewereindependentofthemethodofBPmeasurement.Thisfindingsu pportstheinferencethat office,home,andambulatorydaytimeandnighttimeBP responsesmeasurethesameBPresponsesignal. Oursecondobjectivewastoestimatesignal-to-noise ratiosandcomparethepowertoidentifyageneticpolymorphismpredictingBPresponsewhenmeasuredby eachmethodseparatelyandbyweightedaveragesof multiplemethods.EstimationoftheBPresponsesignal allowedustoalsocompareeachmethodwithatheoreticallyperfectmeasurementconsistingofpuresignal andnonoise.Wereasonedthatgreatersignal-to-noise ratioswouldtranslateintogreaterpowerandsmaller samplesizesrequiredtoidentifyapolymorphisminfluencingBPresponsetoantihypertensivedrugtherapy. Notsurprisingly,signal-to-noiseratiosweregreaterfor thehomeandambulatorydaytimemethods,whichare basedonmoremeasurementspersubjectandhave smallererrorvariances,thanfortheofficeandambulatorynighttimemethods,whicharebasedonfewermeasurementspersubjectandhavelargererrorvariances. Particularlyunsett lingwerethesignal-to-noiseratios lessthanoneforofficeandambulatorynighttimeBP responses,indicatingmore noisethansignalforthese methods.Suchmeasurementimprecisioncouldaccount forlimitedsuccessinpreviousstudiestoidentifypredictorsofofficeBPresponse[12,23]andtherequirement forsamplesizesinthetensofthousandsforgenomewideassociationanalysesofBPlevel[24].Moreover,the profoundlackofpowertoidentifyageneticpredictorof BPresponseinsamplesizes 300,andthelarge increasesinsamplesizerequiredtomaintain80% power,emphasizestheneedformoreprecisemethods ofmeasuringBPresponsethanofficeBPmeasurements provide[25]. AlthoughthehomeandambulatorydaytimeBP responsesprovidedgreaterpowerthantheofficeand ambulatorynighttimeBPresponses,theestimatedsamplesizerequiredtomaintain80%powerwitheither methodwasstillinexcessofthenumberofparticipants randomizedtoeachsingle-drugtherapyinthePEAR study(i.e.,N=400).Consequently,wepursuedadditionalstrategiestoincreasepowerbycombiningallof themeasurementsfrommultiplemethodsinaweighted average,withtheweightschosentominimizetheerror variance(noise)andmaximizethesignal-to-noiseratio oftheresultingaverage.Weprovidedtwoexamples:a weightedaverageofmeasurementsfromallfourmethodsandaweightedaverageofthehomeandofficemeasurements.Thelatterusesthetwomostfeasibleand widelyavailablemethodsofmeasuringantihypertensive drugresponses.Whilebothweightedaveragesdemonstratedimprovementsinthesignal-to-noiseratiosrelativetotheseparatemethodsofmeasuringBPresponse, onlywiththeweightedaverageofallfourmethodswas powermaintainedat80%withoutanincreaseinsample sizeexceedingtheN=400randomizedtoeachsingledrugtherapyinthePEARstudy.Thesesignalandnoise analyses,powercalculations,andsamplesizeestimates basedonthePEARstudyemphasizethe make-orbreak contributionthatprecisioninmeasurementof thephenotypecanmaketosuccessofgenome-wide associationstudies[26]. GivenourinterestinthePEARstudytoidentifynew predictorsofBPresponse,severaladditionalresultsof ouranalysesarenoteworthy.F irst,althoughsignalvariancesweregreaterforthesystolicthanthediastolicBP responses,theerrorvarianceswerealsogreaterandthe signal-to-noiseratiosdifferedlittlebetweenthesystolic anddiastolicBPresponses.Thisfindingsuggeststhat neitherphenotypeaffordsg reateropportunitythanthe othertoidentifyitspredictors.Thissuggestionissupportedbythefindingthateachknownpredictorwasa statisticallysignificantpredictorofbothsystolic and diastolicBPresponses(Table3).Second,greatersignaland signal-to-noiseratiosfor theBPresponsestoatenolol mightsuggestgreaterpredictabilityofBPresponseto atenololthantohydrochlorothiazide.Thissuggestionis supportedbythefindingthattwoidentifiedpredictors ofBPresponsetoatenololwerenotpredictorsofBP responsetohydrochlorothiazide(Table3).Third,male sexwasassociatedwithlessersystolicanddiastolicBP responsestobothatenololandhydrochlorothiazidein thisstudyandtohydrochlorothiazideinaprevious pharmacogeneticstudy[12].Malesexwasalsopreviouslyassociatedwithlesserresponsestoquinapril[27] andcandesartan[13].Toourknowledge,thisapparently consistentassociationofmalesexwithlesserBP responsetodrugsfromdifferentpharmacologicalclasses hasnotbeenpreviouslyrecognized. Despitemanystudiesofantihypertensivedrugsconductedsincethe1950s,fewpatientcharacteristicshave beenidentifiedthatpredictinter-individualdifferences inBPresponses.Methodsthatreducetheerrorinmeasuringbloodpressurerespo nse,especiallyweighted averagesoftheresponsesmeasuredbymultiplemethods,improvesignal-to-noiseratiosandprovidegreater powertoidentifythepredic torsofresponseinsmaller samplesizes.TheirincorporationinthedesignofpharmacogenomicstudiessuchasthePEARstudywillbe criticaltosuccessinidentifyingnovelgeneticTurner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page7of9

PAGE 8

polymorphismsthatimprovetheabilitytopredictBP responsetoantihypertensivedrugtherapy.ConclusionSincedifferentmethodsofmeasuringBPresponseto antihypertensivedrugtherapymeasurethesamesignal, weightedaveragesoftheBPresponsesmeasuredby multiplemethodsminimizemeasurementerrorand optimizepowertoidentifygeneticpredictorsofBP response.AdditionalmaterialAdditionalfile1:Additionalfile1:TableS1 .Esimatedweightsfor calculationofminimumvarianceweightedaveragebloodpressure responses. Acknowledgements Wegratefullyacknowledgethevaluablecontributionsofthestudy participants,supportstaff,andstudyphysicians:Drs.GeorgeBaramidze,R. WhitCurry,KarenHall,KarenHall,FredericRabari-Oskoui,DanRubin,and SiegfriedSchmidt.Inaddition,wegratefullyacknowledgethebiostatistical analysesofDanielCrusan.FundingwasfromNIHPharmacogenetics ResearchNetworkgrantU01-GM074492;K23grantsHL091120(A.L. Beitelshees)andHL086558(R.M.Cooper-DeHoff);CTSAgrantsUL1RR092890(UniversityofFlorida),UL1-RR025008(EmoryUniversity),andUL1RR024150(MayoClinic);andfundsfromtheMayoFoundation. Authordetails1DivisionofNephrologyandHypertension,DepartmentofMedicine,Mayo Clinic,Rochester,MN55905,USA.2DivisionofBiostatistics,Departmentof HealthSciencesResearch,MayoClinic,200FirstStreetS.W,Rochester,MN, USA.3RenalDivision,DepartmentofMedicine,EmoryUniversitySchoolof Medicine,Atlanta,Georgia,USA.4DepartmentofMedicine,Universityof MarylandSchoolofMedicine,Baltimore,MD,USA.5Departmentof PharmacotherapyandTranslationalResearchandCenterfor Pharmacogenomics,UniversityofFlorida,Gainesville,FL,USA.6Department ofCommunityHealthandFamilyMedicine,UniversityofFlorida,Gainesville, FL,USA.7DepartmentofMedicine,UniversityofFlorida,Gainesville,FL,USA.8HumanGeneticsandInstituteofMolecularMedicine,UniversityofTexas HealthScienceCenter,Houston,TX,USA. Authors contributions STandGSparticipatedinthedesignofthestudy,collectedandanalyzed thedata,anddraftedandrevisedthemanuscript.AB,JG,RCDparticipated inthedesignofthestudy,collectedthedataandparticipatedinrevisionof themanuscript.JJconceivedanddesignedthestudy,collectedthedata, andparticipatedinrevisionofthemanuscript.EBparticipatedindesignof thestudyandrevisionofthemanuscript.KBparticipatedindesignofthe study,analyzedthedata,andparticipatedindraftingandrevisionofthe manuscript.Allauthorsreadandapprovedthefinalmanuscript Competinginterests Theauthorsdeclarethattheyhavenocompetinginterests. Received:20December2011Accepted:13March2012 Published:13March2012 References1.ParatiG,OmboniS,BiloG: WhyIsOut-of-OfficeBloodPressure MeasurementNeeded?HomeBloodPressureMeasurementsWill IncreasinglyReplaceAmbulatoryBloodPressureMonitoringinthe DiagnosisandManagementofHypertension. Hypertension 2009, 54 :181-187. 2.AppelLJ,StasonWB: Ambulatorybloodpressuremonitoringandblood pressureself-measurementinthediagnosisandmanagementof hypertension. AnnInternMed 1993, 118 :867-882. 3.ManciaG,ParatiG: Ambulatorybloodpressuremonitoringandorgan damage. Hypertension 2000, 36 :894-900. 4.BobrieG,ChatellierG,GenesN,ClersonP,VaurL,VaisseB,MenardJ, MallionJM: Cardiovascularprognosisof maskedhypertension detected bybloodpressureself-measurementinelderlytreatedhypertensive patients. JAMA 2004, 291 :1342-1349. 5.StaessenJA,DenHondE,CelisH,FagardR,KearyL,VandenhovenG, O BrienET: Antihypertensivetreatmentbasedonbloodpressure measurementathomeorinthephysician soffice:arandomized controlledtrial. JAMA 2004, 291 :955-964. 6.StaessenJA,ThijsL,FagardR,O BrienET,ClementD,deLeeuwPW, ManciaG,NachevC,PalatiniP,ParatiG, etal : Predictingcardiovascular riskusingconventionalvsambulatorybloodpressureinolderpatients withsystolichypertension.SystolicHypertensioninEuropeTrial Investigators[seecomments]. JAMA 1999, 282 :539-546. 7.ManciaG,ParatiG: Officecomparedwithambulatorybloodpressurein assessingresponsetoantihypertensivetreatment:ameta-analysis. J Hypertens 2004, 22 :435-445. 8.IshikawaJ,CarrollDJ,KuruvillaS,SchwartzJE,PickeringTG: Changesin homeversusclinicbloodpressurewithantihypertensivetreatments:a meta-analysis. Hypertension 2008, 52 :856-864. 9.TurnerST,SchwartzGL,BoerwinkleE: Personalizedmedicineforhigh bloodpressure. Hypertension 2007, 50 :1-5. 10.PrestonRA,MatersonBJ,RedaDJ,WilliamsDW,HamburgerRJ, CushmanWC,AndersonRJ: Age-racesubgroupcomparedwithrenin profileaspredictorsofbloodpressureresponsetoantihypertensive therapy.DepartmentofVeteransAffairsCooperativeStudyGroupon AntihypertensiveAgents. JAMA 1998, 280 :1168-1172. 11.TurnerST,SchwartzGL,ChapmanAB,BeitelsheesAL,GumsJG,CooperDeHoffRM,BoerwinkleE,JohnsonJA,BaileyKR: Plasmareninactivity predictsbloodpressureresponsestobeta-blockerandthiazidediuretic asmonotherapyandadd-ontherapyforhypertension. AmJHypertens 2010, 23 :1014-1022. 12.ChapmanAB,SchwartzGL,BoerwinkleE,TurnerST: Predictorsof antihypertensiveresponsetoastandarddoseofhydrochlorothiazidefor essential hypertension. KidneyInt 2002, 61 :1047-1055. 13.CanzanelloVJ,Baranco-PryorE,Rahbari-OskouiF,SchwartzGL,BoerwinkleE, TurnerST,ChapmanAB: Predictorsofbloodpressureresponsetothe angiotensinreceptorblockercandesartaninessentialhypertension. Am JHypertens 2008, 21 :61-66. 14.FinkielmanJD,SchwartzGL,ChapmanAB,BoerwinkleE,TurnerST: Reproducibilityofbloodpressureresponsetohydrochlorothiazide. JClin Hypertens(Greenwich) 2002, 4 :408-412. 15.FinkielmanJD,SchwartzGL,ChapmanAB,BoerwinkleE,TurnerST: Lackof agreementbetweenofficeandambulatorybloodpressureresponsesto hydrochlorothiazide. AmJHypertens 2005, 18 :398-402. 16.BeitelsheesAL,GongY,BaileyKR,TurnerST,ChapmanAB,SchwartzGL, GumsJG,BoerwinkleE,JohnsonJA: Comparisonofoffice,ambulatory, andhomebloodpressureantihypertensiveresponsetoatenololand hydrochlorthiazide. JClinHypertens(Greenwich) 2010, 12 :14-21. 17.JohnsonJA,BoerwinkleE,ZinehI,ChapmanAB,BaileyK,CooperDeHoffRM,GumsJ,CurryRW,GongY,BeitelsheesAL, etal : Pharmacogenomicsofantihypertensivedrugs:rationaleanddesignof thePharmacogenomicEvaluationofAntihypertensiveResponses(PEAR) study. AmHeartJ 2009, 157 :442-449. 18.TopouchianJA,ElAssaadMA,OrobinskaiaLV,ElFeghaliRN,AsmarRG: Validationoftwodevicesforself-measurementofbrachialblood pressureaccordingtotheInternationalProtocoloftheEuropeanSociety ofHypertension:theSEINEXSE-9400andtheMicrolifeBP3AC1-1. Blood PressMonit 2005, 10 :325-331. 19.O BrienE,MeeF,AtkinsN,O MalleyK: AccuracyoftheSpaceLabs90207 determinedbytheBritishHypertensionSocietyprotocol. JHypertens 1991, 9 :573-574. 20.GillJS,ZezulkaAV,BeeversDG,DaviesP: Relationbetweeninitialblood pressureanditsfallwithtreatment. Lancet 1985, 1 :567-569. 21.SnedecorGW,CochranWG: Statisticalmethods. 7edition.AmesIA,:Iowa StateUniv.Press;1980.Turner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page8of9

PAGE 9

22.RischN,MerikangasK: Thefutureofgeneticstudiesofcomplexhuman diseases. Science 1996, 273 :1516-1517. 23.CannellaG,PaolettiE,BarocciS,MassarinoF,DelfinoR,RaveraG,Di MaioG,NoceraA,PatroneP,RollaD: Angiotensin-convertingenzyme genepolymorphismandreversibilityofuremicleftventricular hypertrophyfollowinglong-termantihypertensivetherapy. KidneyInt 1998, 54 :618-626. 24.LevyD,EhretGB,RiceK,VerwoertGC,LaunerLJ,DehghanA,GlazerNL, MorrisonAC,JohnsonAD,AspelundT, etal : Genome-wideassociation studyofbloodpressureandhypertension. NatGenet 2009, 41 :677-687. 25.BellKJ,HayenA,MacaskillP,CraigJC,NealBC,FoxKM,RemmeWJ, AsselbergsFW,vanGilstWH,MacmahonS, etal : Monitoringinitial responsetoAngiotensin-convertingenzymeinhibitor-basedregimens: anindividualpatientdatameta-analysisfromrandomized,placebocontrolledtrials. Hypertension 2010, 56 :533-539. 26.MacraeCA,VasanRS: Next-generationgenome-wideassociationstudies: timetofocusonphenotype? CircCardiovascGenet 2011, 4 :334-336. 27.MokweE,OhmitSE,NasserSA,ShafiT,SaundersE,CrookE,DudleyA, FlackJM: Determinantsofbloodpressureresponsetoquinaprilinblack andwhitehypertensivepatients:theQuinaprilTitrationInterval ManagementEvaluationtrial. Hypertension 2004, 43 :1202-1207.doi:10.1186/1479-5876-10-47 Citethisarticleas: Turner etal .: Powertoidentifyageneticpredictorof antihypertensivedrugresponseusingdifferentmethodstomeasure bloodpressureresponse. JournalofTranslationalMedicine 2012 10 :47. Submit your next manuscript to BioMed Central and take full advantage of: Convenient online submission Thorough peer review No space constraints or color gure charges Immediate publication on acceptance Inclusion in PubMed, CAS, Scopus and Google Scholar Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit Turner etal JournalofTranslationalMedicine 2012, 10 :47 http://www.translational-medicine.com/content/10/1/47 Page9of9

PAGE 10

Supplementary Table S 1. Esimated weights for calculation of minimum variance weighted average blood pressure responses BP Response to Atenolol BP Response to Hydrochlorothiazide Systolic Diastolic Systolic Diastolic Average of all four methods Home 0.45 0.48 0.45 0.48 Ambulatory daytime 0.40 0.35 0.31 0.27 Ambulatory nighttime 0.15 0.10 0.18 0.16 Office 0.00 0.07 0.05 0.09 Average of three methods Home 0.50 0.51 0.49 0.52 Ambulatory daytime 0.49 0.42 0.44 0.36 Office 0. 01 0.07 0.07 0.12 Average of two methods Home 0.84 0.79 0.81 0.74 Office 0.16 0.21 0.19 0.26 The weighted average combinations were determined based on the row sums of the inverse of the inter method covariance matrices, which provide weights t hat minimize the variance (see Methods)


xml version 1.0 encoding utf-8 standalone no
mets ID sort-mets_mets OBJID sword-mets LABEL DSpace SWORD Item PROFILE METS SIP Profile xmlns http:www.loc.govMETS
xmlns:xlink http:www.w3.org1999xlink xmlns:xsi http:www.w3.org2001XMLSchema-instance
xsi:schemaLocation http:www.loc.govstandardsmetsmets.xsd
metsHdr CREATEDATE 2012-05-02T16:09:49
agent ROLE CUSTODIAN TYPE ORGANIZATION
name BioMed Central
dmdSec sword-mets-dmd-1 GROUPID sword-mets-dmd-1_group-1
mdWrap SWAP Metadata MDTYPE OTHER OTHERMDTYPE EPDCX MIMETYPE textxml
xmlData
epdcx:descriptionSet xmlns:epdcx http:purl.orgeprintepdcx2006-11-16 xmlns:MIOJAVI
http:purl.orgeprintepdcxxsd2006-11-16epdcx.xsd
epdcx:description epdcx:resourceId sword-mets-epdcx-1
epdcx:statement epdcx:propertyURI http:purl.orgdcelements1.1type epdcx:valueURI http:purl.orgeprintentityTypeScholarlyWork
http:purl.orgdcelements1.1title
epdcx:valueString Power to identify a genetic predictor of antihypertensive drug response using different methods to measure blood pressure response
http:purl.orgdctermsabstract
Abstract
Background
To determine whether office, home, ambulatory daytime and nighttime blood pressure (BP) responses to antihypertensive drug therapy measure the same signal and which method provides greatest power to identify genetic predictors of BP response.
Methods
We analyzed office, home, ambulatory daytime and nighttime BP responses in hypertensive adults randomized to atenolol (N = 242) or hydrochlorothiazide (N = 257) in the Pharmacogenomic Evaluation of Antihypertensive Responses Study. Since different measured BP responses may have different predictors, we tested the "same signal" model by using linear regression methods to determine whether known predictors of BP response depend on the method of BP measurement. We estimated signal-to-noise ratios and compared power to identify a genetic polymorphism predicting BP response measured by each method separately and by weighted averages of multiple methods.
Results
After adjustment for pretreatment BP level, known predictors of BP response including plasma renin activity, race, and sex were independent of the method of BP measurement. Signal-to-noise ratios were more than 2-fold greater for home and ambulatory daytime BP responses than for office and ambulatory nighttime BP responses and up to 11-fold greater for weighted averages of all four methods. Power to identify a genetic polymorphism predicting BP response was directly related to the signal-to-noise ratio and, therefore, greatest with the weighted averages.
Conclusion
Since different methods of measuring BP response to antihypertensive drug therapy measure the same signal, weighted averages of the BP responses measured by multiple methods minimize measurement error and optimize power to identify genetic predictors of BP response.
http:purl.orgdcelements1.1creator
Turner, Stephen T
Schwartz, Gary L
Chapman, Arlene B
Beitelshees, Amber L
Gums, John G
Cooper-DeHoff, Rhonda M
Boerwinkle, Eric
Johnson, Julie A
Bailey, Kent R
http:purl.orgeprinttermsisExpressedAs epdcx:valueRef sword-mets-expr-1
http:purl.orgeprintentityTypeExpression
http:purl.orgdcelements1.1language epdcx:vesURI http:purl.orgdctermsRFC3066
en
http:purl.orgeprinttermsType
http:purl.orgeprinttypeJournalArticle
http:purl.orgdctermsavailable
epdcx:sesURI http:purl.orgdctermsW3CDTF 2012-03-13
http:purl.orgdcelements1.1publisher
BioMed Central Ltd
http:purl.orgeprinttermsstatus http:purl.orgeprinttermsStatus
http:purl.orgeprintstatusPeerReviewed
http:purl.orgeprinttermscopyrightHolder
Turner et al.; licensee BioMed Central Ltd.
http:purl.orgdctermslicense
http://creativecommons.org/licenses/by/2.0
http:purl.orgdctermsaccessRights http:purl.orgeprinttermsAccessRights
http:purl.orgeprintaccessRightsOpenAccess
http:purl.orgeprinttermsbibliographicCitation
Journal of Translational Medicine. 2012 Mar 13;10(1):47
http:purl.orgdcelements1.1identifier
http:purl.orgdctermsURI http://dx.doi.org/10.1186/1479-5876-10-47
fileSec
fileGrp sword-mets-fgrp-1 USE CONTENT
file sword-mets-fgid-0 sword-mets-file-1
FLocat LOCTYPE URL xlink:href 1479-5876-10-47.xml
sword-mets-fgid-1 sword-mets-file-2 applicationpdf
1479-5876-10-47.pdf
sword-mets-fgid-3 sword-mets-file-3 applicationmsword
1479-5876-10-47-S1.DOC
structMap sword-mets-struct-1 structure LOGICAL
div sword-mets-div-1 DMDID Object
sword-mets-div-2 File
fptr FILEID
sword-mets-div-3
sword-mets-div-4


!DOCTYPE art SYSTEM 'http:www.biomedcentral.comxmlarticle.dtd'
ui 1479-5876-10-47
ji 1479-5876
fm
dochead Research
bibl
title p Power to identify a genetic predictor of antihypertensive drug response using different methods to measure blood pressure response
aug
au id A1 ca yes snm Turnermi Tfnm Stepheninsr iid I1 email sturner@mayo.edu
A2 SchwartzLGarygschwartz@mayo.edu
A3 ChapmanBArleneI3 Arlene_Chapman@emory.org
A4 BeitelsheesLAmberI4 abeitels@medicine.umaryland.edu
A5 GumsGJohnI5 I6 jgums@ufl.edu
A6 Cooper-DeHoffMRhondaI7 dehoff@cop.ufl.edu
A7 BoerwinkleEricI8 Eric.Boerwinkle@uth.tmc.edu
A8 JohnsonAJuliejohnson@cop.ufl.edu
A9 BaileyRKentI2 baileyk@mayo.edu
insg
ins Division of Nephrology and Hypertension, Department of Medicine, Mayo Clinic, Rochester, MN 55905, USA
Division of Biostatistics, Department of Health Sciences Research, Mayo Clinic, 200 First Street S.W, Rochester, MN, USA
Renal Division, Department of Medicine, Emory University School of Medicine, Atlanta, Georgia, USA
Department of Medicine, University of Maryland School of Medicine, Baltimore, MD, USA
Department of Pharmacotherapy and Translational Research and Center for Pharmacogenomics, University of Florida, Gainesville, FL, USA
Department of Community Health and Family Medicine, University of Florida, Gainesville, FL, USA
Department of Medicine, University of Florida, Gainesville, FL, USA
Human Genetics and Institute of Molecular Medicine, University of Texas Health Science Center, Houston, TX, USA
source Journal of Translational Medicine
section Clinical translationissn 1479-5876
pubdate 2012
volume 10
issue 1
fpage 47
url http://www.translational-medicine.com/content/10/1/47
xrefbib pubidlist pubid idtype doi 10.1186/1479-5876-10-47pmpid 22413836
history rec date day 20month 12year 2011acc 1332012pub 1332012cpyrt 2012collab Turner et al; licensee BioMed Central Ltd.note This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
kwdg
kwd hypertensionblood pressure monitoringantihypertensive drug therapybeta-blockerthiazide diureticplasma renin activity
abs
sec st Abstract
Background
To determine whether office, home, ambulatory daytime and nighttime blood pressure (BP) responses to antihypertensive drug therapy measure the same signal and which method provides greatest power to identify genetic predictors of BP response.
Methods
We analyzed office, home, ambulatory daytime and nighttime BP responses in hypertensive adults randomized to atenolol (N = 242) or hydrochlorothiazide (N = 257) in the Pharmacogenomic Evaluation of Antihypertensive Responses Study. Since different measured BP responses may have different predictors, we tested the "same signal" model by using linear regression methods to determine whether known predictors of BP response depend on the method of BP measurement. We estimated signal-to-noise ratios and compared power to identify a genetic polymorphism predicting BP response measured by each method separately and by weighted averages of multiple methods.
Results
After adjustment for pretreatment BP level, known predictors of BP response including plasma renin activity, race, and sex were independent of the method of BP measurement. Signal-to-noise ratios were more than 2-fold greater for home and ambulatory daytime BP responses than for office and ambulatory nighttime BP responses and up to 11-fold greater for weighted averages of all four methods. Power to identify a genetic polymorphism predicting BP response was directly related to the signal-to-noise ratio and, therefore, greatest with the weighted averages.
Conclusion
Since different methods of measuring BP response to antihypertensive drug therapy measure the same signal, weighted averages of the BP responses measured by multiple methods minimize measurement error and optimize power to identify genetic predictors of BP response.
bdy
Background
Although office blood pressure (BP) measurements remain the standard-of-care, averages of out-of-office measurements are more reproducible abbrgrp abbr bid B1 1. Out-of-office averages have also been reported to be more strongly correlated with subclinical target organ damage B2 2B3 3 and to better predict future cardiovascular disease events B4 4B5 5B6 6 than office measurements. Not surprisingly, BP responses to antihypertensive drug therapy are more precisely and accurately determined by out-of-office than office measurements, which are influenced by white coat and placebo effects B7 7B8 8. Consequently, greater use of out-of-office methods of BP measurement has been advocated for clinical decision-making and research 1.
More individualized approaches to antihypertensive drug therapy may become possible if genetic polymorphisms are discovered that improve the ability to predict inter-individual differences in BP response B9 9. Known predictors are limited to race, age, and plasma renin activity B10 10B11 11, which explain less than 50% of interindividual variation in BP response to single-drug therapy B12 12B13 13. Most previous studies have attempted to identify genetic or non-genetic predictors of office BP response, which is not very reproducible and correlates only modestly with home and ambulatory BP responses 78B14 14B15 15B16 16. Whether out-of-office measurements of BP response can improve the ability to identify predictors of BP response has not been demonstrated. Method-specific measurement errors could account for differences in the magnitude of and correlation between office, home, ambulatory daytime and nighttime BP responses 8. However, an additional possibility is that different BP response signals are measured by the different methods.
Since different BP response signals may have different predictors, our first objective in the present study was to test the "same signal" model by determining whether known predictors of BP response, i.e., race, age, and plasma renin activity, depend on the method of BP measurement. We analyzed data from the Pharmacogenomic Evaluation of Antihypertensive Responses (PEAR) study, in which BP responses to single-drug therapy with atenolol or hydrochlorothiazide were measured by all four methods 78. In this context, our second objective was to estimate signal-to-noise ratios and compare the power to identify a genetic polymorphism predicting BP response when measured by each method separately and by weighted averages of multiple methods.
Methods
Participants
The PEAR study B17 17 http://clinicaltrials.gov/ct2/show/NCT00246519 was approved by the Institutional Review Board at each site, and all participants gave informed consent. At an initial consent and screening visit, trained study personnel administered standardized questionnaires, performed a limited physical examination, and obtained blood and urine samples for testing to establish eligibility for participation 11. Participants were provided an automated sphygmomanometer (MicroLife 3 AC1-PC, Minneapolis MN), the adequacy of which has been previously validated B18 18, and withdrawn from previous antihypertensive drug therapy. The device was set to measure BP in triplicate with each activation and to store the average systolic and diastolic BPs and the time of each set of measurements. Participants were instructed to take readings daily in the seated position, one set of three readings in the morning upon arising from bed and a second set in the evening just before retiring. At subsequent study visits (prior to randomization and at the end of therapy), an additional set of three readings was obtained seated (> 5 minutes) in the office using the home monitor. In addition, 24-hour ambulatory BP recordings were obtained at these visits using Spacelabs (Redmond WA) ambulatory monitors, model 90207, the adequacy of which has been previously validated B19 19. Participants were instructed to conduct their usual daily activities while wearing the monitor, which was set to record BP four times per hour during the day (6 AM to 10 PM) and twice per hour during the night (10 PM to 6 AM). The average (± standard deviation) number of ambulatory measurements was 67 ± 10 during daytime hours and 15 ± 3 during nighttime hours.
At the end of the drug-free washout period, fasting blood samples were drawn in the seated position after ambulation for measurement of plasma renin activity 11. To qualify for randomization, the average home diastolic BP in the previous week had to be ≥85 mmHg (consisting of at least five morning and five evening sets of readings) it and the average office diastolic BP ≥ 90 mmHg. Participants received either atenolol or hydrochlorothiazide, starting at 50 mg or 12.5 mg daily, respectively, for two weeks, after which, if BP remained > 120/70 mmHg, the doses were increased to 100 mg or 25 mg daily, respectively, for six additional weeks.
Statistical analysis
Analyses were performed with Statistical Analysis System software, version 9.1 (SAS, Raleigh-Durham NC). Statistical significance was defined a priori by P < 0.05. The BP response to each drug was calculated for each measurement method by subtracting the pretreatment average from the post-treatment average. The home BP averages consisted of at least five of seven morning and evening sets of three readings taken during the week prior to the pre- and post-treatment study visits (i.e., at least 30 and up to 42 measurements prior to each visit). Multiple-variable linear regression analyses were performed to identify participant characteristics that made additive, statistically independent contributions to the prediction of systolic and diastolic BP response to each drug. In preliminary analyses, we found that higher pretreatment BP level was associated with greater BP response, as expected B20 20. Because we sought to evaluate predictors that are independent of the pretreatment BP level, we first regressed out the effects of pretreatment BP level and then modeled the effects of other known predictors of BP response 1011 as well as other variables measured at the consent and screening visit 1117. Final multiple-variable models were derived using a backward stepwise elimination procedure, retaining only the predictors of both systolic and diastolic BP responses to either drug. In the initial models that included race and pretreatment plasma renin activity, age was not a statistically significant predictor of BP responses and was not retained in the final models.
To determine whether known predictors of BP response depend on the method of measuring BP response, we compared models in which regression coefficients were constrained to be identical among measurement methods or allowed to differ among methods by including interactions of each predictor with the method of BP measurement. Considering the model in which the regression coefficients were identical across methods as a null hypothesis, we attempted to detect any departures indicating dependency of the predictors on method of BP measurement that would lead us to reject the "same signal" model. This analysis used PROC GENMOD in SAS, which adjusts for the correlation among the four BP response measurements within each participant.
To estimate signal-to-noise ratios, the covariance matrix of the four measured BP responses was used to estimate the signal and noise components for each method of measuring BP response after regressing out the method-specific effects of pretreatment BP level. The correlation coefficient between BP responses measured by two methods provides a dimensionless measure of how much the two responses covary (change together); the covariance between them expresses the correlation in units of the two BP responses multiplied together (mmsup 2Hg) and is the variance shared between them, i.e., the signal variance. Since each pair-wise covariance provides an unbiased estimate of the signal variance, we used the average of the six pairwise covariances as the BP response signal. Subtracting the signal from the method-specific total variance provided an estimate of the method-specific error variance or noise.
We examined implications of the signal-to-noise analyses for accomplishing the goal of the PEAR and other pharmacogenomic studies. Specifically, we compared power and samples sizes required to identify a genetic polymorphism that predicts BP response when measured by each method separately and by weighted averages of the responses measured by multiple methods. The rationale for the weighted averages was to increase the signal-to-noise ratio (and power) by minimizing the error variance. Two different combinations of the measured BP responses were considered: a weighted average of all four methods and a weighted average of the office and home BP responses. The weighted average combinations were determined based on the row sums of the inverse of the inter-method covariance matrices, which provide weights that minimize the variance B21 21.
For the power and samples size calculations, we assumed that a genetic polymorphism with a minor allele frequency of 0.2 influences the BP response signal with an effect size that can be detected with 80% power in a sample of N = 300 at a genome-wide significance level of 5 × 10-8. This P-value was originally suggested for genome-wide association analysis of 1 million single nucleotide polymorphisms using a Bonferroni correction for multiple testing B22 22. Based on the estimated signal variances, we calculated the allele effect sizes (in mmHg/allele) and the percentage of variation in the BP response (R2×100%) explained by the polymorphism. We then calculated the power to detect the polymorphism in a sample of N = 300 when the BP response is measured by each method separately and weighted averages of multiple methods, and the corresponding sample sizes required to maintain 80% power.
Results
Sample description
Five hundred and ninety-five study participants had complete measurements of office, home, and ambulatory daytime and nighttime BP responses (Table tblr tid T1 1). Of these, 293 participants were randomized to atenolol (49%) and 302 to hydrochlorothiazide (51%) treatment. Mean values and relative frequencies of participant characteristics measured prior to randomization did not differ significantly between the atenolol and hydrochlorothiazide-treated groups 11 (not shown).
tbl Table 1caption Descriptive characteristics of study participantstblbdy cols 2
r
c
left
b Mean ± standard deviation or N (%)
cspan
hr
N (%)
595 (100)
Randomized to hydrochlorothiazide, N (%)
302 (51)
Age, years
49.3 ± 9.1
Male, N (%)
280 (47)
Black, N (%)
245 (41)
BMI, kg·m-2
30.6 ± 5.6
Hypertension duration, years
7.1 ± 7.2
Antihypertensive medication, N (%)
492 (91)
Current smoker, N (%)
69 (13)
Glucose, mg·dL-1
94.8 ± 10.5
Creatinine, mg·dL-1
0.9 ± 0.2
Serum ALT, U·L-1
29.1 ± 15.9
Plasma renin activity, ng·mL-1·hr-1
1.0 ± 1.2
Resting heart rate, beat·min-1
71.0 ± 10.2
Screening office systolic BP, mmHg
137.9 ± 13.8
Screening office diastolic BP, mmHg
89.5 ± 8.8
Pretreatment systolic blood pressure, mmHg
151.5 ± 13.8
Pretreatment diastolic blood pressure, mmHg
98.2 ± 6.3
tblfn
BMI, body mass index; ALT, alanine aminotransferase; BP, mmHg. Characteristics were measured at the screening visit except pretreatment systolic and diastolic BP and plasma renin activity were measured at the end of the drug-free washout period prior to initiating atenolol or hydrochlorothiazide therapy.
Office, home, and ambulatory BP response
Means and standard deviations of the systolic and diastolic BP responses differed among measurement methods (Table T2 2). For systolic BP response, office measurements had the greatest mean declines and home measurements the smallest mean declines in response to each drug (Table 2). For diastolic BP response, office measurements also had the greatest mean declines in response to each drug; ambulatory nighttime measurements had the smallest mean decline in response to atenolol and home measurements the smallest mean decline in response to hydrochlorothiazide. Correlation coefficients between the office, home, ambulatory daytime and nighttime BP responses were modest in magnitude (not shown), ranging from 0.36 to 0.71 after adjustment for differences in pretreatment BP levels (all P < 0.0001).
Table 2Blood Pressure Responses to Monotherapy by Measurement Method4
center
All N = 595
Atenolol N = 293
Hydrochlorothiazide N = 302
right
Systolic BP Response, mmHg
Office
-13.4 ± 14.7
-13.5 ± 15.6
-13.2 ± 13.7
Home
-8.8 ± 9.8
-8.3 ± 10.4
-9.4 ± 9.1
Ambulatory daytime
-11.5 ± 10.5
-12.2 ± 11.1
-10.8 ± 9.8
Ambulatory nighttime
-9.7 ± 12.5
-8.9 ± 12.9
-10.6 ± 12.1
Contrast P value
< 0.001
< 0.001
< 0.001
Diastolic BP Response, mmHg
Office
-8.6 ± 8.9
-10.5 ± 9.4
-6.8 ± 7.9
Home
-6.6 ± 6.5
-7.8 ± 6.8
-5.3 ± 6.0
Ambulatory daytime
-7.6 ± 7.6
-9.2 ± 7.9
-6.1 ± 7.0
Ambulatory nighttime
-6.8 ± 9.5
-7.0 ± 10.0
-6.5 ± 9.0
Contrast P value
< 0.001
< 0.001
0.10
BP, blood pressure.
The BP response signal and its predictors
We assessed whether the office, home, and ambulatory daytime and nighttime BP responses measure the same BP response signal by determining whether the predictors of BP response depend on the method of BP measurement (see Methods). After adjustment for pretreatment BP level, none of the predictors of BP response depended upon of the method of BP measurement (analyses not shown). For all four methods of measuring BP response, the identified predictors included race, plasma renin activity, and sex (Table T3 3). As expected, black race was associated with lesser systolic and diastolic BP responses to atenolol and greater responses to hydrochlorothiazide; and greater log renin was associated with greater systolic and diastolic BP responses to atenolol and lesser responses to hydrochlorothiazide 11. Male sex was independently associated with lesser systolic and diastolic BP responses to each drug. Greater log hypertension years and greater serum ALT were each independently associated with greater systolic and diastolic BP responses to atenolol but not to hydrochlorothiazide.
Table 3Multi-variable linear regression modeling of predictors of blood pressure response signal after adjustment for pretreatment blood pressure level5
BP Response to Atenolol (N = 293)
BP Response to Hydrochlorothiazide (N = 302)
Systolic β ± SE
Diastolic β ± SE
Systolic β ± SE
Diastolic β ± SE
Intercept
-14.1 ± 0.9§
-11.8 ± 0.6§
-11.8 ± 0.7§
-6.7 ± 0.5§
Race: Black
6.3 ± 1.1§
4.8 ± 0.8§
-2.9 ± 0.9‡
-2.0 ± 0.6‡
Log Plasma Renin Activity
-4.4 ± 0.5§
-2.5 ± 0.3§
0.7 ± 0.5
0.7 ± 0.3*
Sex: Male
2.3 ± 1.0*
2.9 ± 0.7§
4.2 ± 0.8§
2.7 ± 0.5§
Log Hypertension Years
1.3 ± 0.5*
1.0 ± 0.4†
-
-
Serum ALT
0.08 ± 0.03†
0.04 ± 0.02*
-
-
Model R2×100%
22%
23%
11%
16%
BP, blood pressure; β, regression coefficient; SE, standard error. Model parameters are estimated at the mean values for each quantitative predictor variable in a combined dataset used to model the predictors of office, home, ambulatory daytime and nighttime BP responses after adjustment for pretreatment BP levels. P-values for tests of model parameters = 0: *, ≤0.05; †, ≤0.01; ‡, ≤ 0.001; §, ≤0.0001. R2×100% is the percentage of variation in the office BP response explained by the model predictors.
Signal-to noise-ratios
Inferring that all four methods measure the same BP response signal, we estimated the signal variance (see Methods) and calculated the method-specific error variance (noise) and signal-to-noise ratio for each measured BP response (Figures figr fid F1 1 and F2 2). The home and ambulatory daytime BP responses had the largest signal-to-noise ratios and the ambulatory nighttime and office BP responses the smallest signal-to-noise ratios (Figure 2). The signal-to-noise ratios of the home and ambulatory daytime BP responses were similar in magnitude and up to 4-fold greater than the signal-to-noise ratios of the office and ambulatory nighttime BP responses, which were mostly less than 1 (more noise than signal). Weighted averages of all four measured BP responses improved the signal-to-noise ratios up to 4-fold compared to the home BP responses and up to 19-fold compared to the office BP responses. Weighted averages of the home and office BP responses improved the signal-to-noise ratios modestly compared to the home BP responses (by 24% at most). The weightings, provided in Additional file supplr sid S1 1: Table S1, minimized the error variance (noise) of the average BP responses (see Methods), thereby accounting for the improvement in signal-to-noise ratios.
fig Figure 1Signal, noise, and total variances of the measured blood pressure responses to single-drug therapy with atenolol or hydrochlorothiazide and their weighted averagestext
Signal, noise, and total variances of the measured blood pressure responses to single-drug therapy with atenolol or hydrochlorothiazide and their weighted averages.
graphic file 1479-5876-10-47-1
Figure 2Signal-to-noise ratios of the measured blood pressure responses to single-drug therapy with atenolol or hydrochlorothiazide and their weighted averages
Signal-to-noise ratios of the measured blood pressure responses to single-drug therapy with atenolol or hydrochlorothiazide and their weighted averages.
1479-5876-10-47-2
suppl
Additional file 1
Additional file 1: Table S1. Esimated weights for calculation of minimum variance weighted average blood pressure responses.
name 1479-5876-10-47-S1.DOC
Click here for file
Power and sample size needed to identify a genetic predictor of blood pressure response
We assumed that a genetic polymorphism with minor allele frequency of 0.2 influences the BP response signal and is detected with 80% power at genome-wide significance level of 5 × 10-8 (see Methods) 22. The T-statistic for association of the polymorphism with BP response is 6.29; 11.6% of the signal variation is explained (R2×100%); and the corresponding effect sizes (β-coefficients) in mm Hg per allele are 5.22/3.36 for the systolic/diastolic BP responses to atenolol and 4.04/2.63 for the systolic/diastolic BP responses to hydrochlorothiazide. Based on the signal and noise analyses, power to detect this polymorphism in a sample size of N = 300 declined for all methods of measuring BP response when compared to a perfect method capable of measuring only signal and no noise (Table T4 4). Power declined most markedly for the office BP responses (to < 5% power) and was only maintained at > 50% for the weighted averages of all four methods. Alternatively, to maintain 80% power the sample sizes would need to be increased for all methods of measuring BP response when compared to the N = 300 sample size for the perfect measurement of signal without noise (Table 4). The sample sizes increased most markedly for the office BP responses (by > 200%), but only modestly for the weighted averages of all four methods (by ≤ 26%).
Table 4Power and sample sizes to detect single nucleotide polymorphism influencing BP response measured by office, home, ambulatory daytime and nighttime blood pressure
Drug
Atenolol
Hydrochlorothiazide
Power, N = 300
N, 80% power
Power, N = 300
N, 80% power
Systolic BP response signal
80%
300
80%
300
indent 1
Measurement methods
Office
4%
780
1.5%
1013
Home
42%
416
24%
500
Ambulatory day
41%
420
18%
540
Ambulatory night
17%
552
12%
607
Weighted averages
All methods
70%
328
61%
353
Home and office
45%
404
30%
467
Diastolic BP response signal
80%
300
80%
300
Measurement methods
Office
5%
764
3%
884
Home
39%
426
19%
537
Ambulatory day
39%
426
19%
537
Ambulatory night
27%
484
7%
695
Weighted averages
All methods
63%
348
379
Home and office
42%
416
26%
489
BP, blood pressure. The power and sample size estimates are for a single nucleotide polymorphism with minor allele frequency of 0.2 that influences the BP response signal with an effect size detected with 80% power in a sample of N = 300 at a genome-wide significance level of 5 × 10-8(see Methods). The estimates assume that the BP response signal can be measured without error.
Discussion
Our first objective was to assess whether office, home, and ambulatory daytime and nighttime measurements of BP response to single-drug therapy measure the same BP response signal. Since different BP response signals may have different predictors, this assessment was based on determining whether known predictors of BP response 11 depend on the method of BP measurement. After adjustment for the method-specific effects of pretreatment BP level, the identified predictors of BP response were independent of the method of BP measurement. This finding supports the inference that office, home, and ambulatory daytime and nighttime BP responses measure the same BP response signal.
Our second objective was to estimate signal-to-noise ratios and compare the power to identify a genetic polymorphism predicting BP response when measured by each method separately and by weighted averages of multiple methods. Estimation of the BP response signal allowed us to also compare each method with a theoretically perfect measurement consisting of pure signal and no noise. We reasoned that greater signal-to-noise ratios would translate into greater power and smaller sample sizes required to identify a polymorphism influencing BP response to antihypertensive drug therapy. Not surprisingly, signal-to-noise ratios were greater for the home and ambulatory daytime methods, which are based on more measurements per subject and have smaller error variances, than for the office and ambulatory nighttime methods, which are based on fewer measurements per subject and have larger error variances. Particularly unsettling were the signal-to-noise ratios less than one for office and ambulatory nighttime BP responses, indicating more noise than signal for these methods. Such measurement imprecision could account for limited success in previous studies to identify predictors of office BP response 12B23 23 and the requirement for sample sizes in the tens of thousands for genome-wide association analyses of BP level B24 24. Moreover, the profound lack of power to identify a genetic predictor of BP response in sample sizes ≤300, and the large increases in sample size required to maintain 80% power, emphasizes the need for more precise methods of measuring BP response than office BP measurements provide B25 25.
Although the home and ambulatory daytime BP responses provided greater power than the office and ambulatory nighttime BP responses, the estimated sample size required to maintain 80% power with either method was still in excess of the number of participants randomized to each single-drug therapy in the PEAR study (i.e., N = 400). Consequently, we pursued additional strategies to increase power by combining all of the measurements from multiple methods in a weighted average, with the weights chosen to minimize the error variance (noise) and maximize the signal-to-noise ratio of the resulting average. We provided two examples: a weighted average of measurements from all four methods and a weighted average of the home and office measurements. The latter uses the two most feasible and widely available methods of measuring antihypertensive drug responses. While both weighted averages demonstrated improvements in the signal-to-noise ratios relative to the separate methods of measuring BP response, only with the weighted average of all four methods was power maintained at 80% without an increase in sample size exceeding the N = 400 randomized to each single-drug therapy in the PEAR study. These signal and noise analyses, power calculations, and sample size estimates based on the PEAR study emphasize the "make-or-break" contribution that precision in measurement of the phenotype can make to success of genome-wide association studies B26 26.
Given our interest in the PEAR study to identify new predictors of BP response, several additional results of our analyses are noteworthy. First, although signal variances were greater for the systolic than the diastolic BP responses, the error variances were also greater and the signal-to-noise ratios differed little between the systolic and diastolic BP responses. This finding suggests that neither phenotype affords greater opportunity than the other to identify its predictors. This suggestion is supported by the finding that each known predictor was a statistically significant predictor of both systolic and diastolic BP responses (Table 3). Second, greater signal and signal-to-noise ratios for the BP responses to atenolol might suggest greater predictability of BP response to atenolol than to hydrochlorothiazide. This suggestion is supported by the finding that two identified predictors of BP response to atenolol were not predictors of BP response to hydrochlorothiazide (Table 3). Third, male sex was associated with lesser systolic and diastolic BP responses to both atenolol and hydrochlorothiazide in this study and to hydrochlorothiazide in a previous pharmacogenetic study 12. Male sex was also previously associated with lesser responses to quinapril B27 27 and candesartan 13. To our knowledge, this apparently consistent association of male sex with lesser BP response to drugs from different pharmacological classes has not been previously recognized.
Despite many studies of antihypertensive drugs conducted since the 1950s, few patient characteristics have been identified that predict inter-individual differences in BP responses. Methods that reduce the error in measuring blood pressure response, especially weighted averages of the responses measured by multiple methods, improve signal-to-noise ratios and provide greater power to identify the predictors of response in smaller sample sizes. Their incorporation in the design of pharmacogenomic studies such as the PEAR study will be critical to success in identifying novel genetic polymorphisms that improve the ability to predict BP response to antihypertensive drug therapy.
Conclusion
Since different methods of measuring BP response to antihypertensive drug therapy measure the same signal, weighted averages of the BP responses measured by multiple methods minimize measurement error and optimize power to identify genetic predictors of BP response.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ST and GS participated in the design of the study, collected and analyzed the data, and drafted and revised the manuscript. AB, JG, RCD participated in the design of the study, collected the data and participated in revision of the manuscript. JJ conceived and designed the study, collected the data, and participated in revision of the manuscript. EB participated in design of the study and revision of the manuscript. KB participated in design of the study, analyzed the data, and participated in drafting and revision of the manuscript. All authors read and approved the final manuscript
bm
ack
Acknowledgements
We gratefully acknowledge the valuable contributions of the study participants, support staff, and study physicians: Drs. George Baramidze, R. Whit Curry, Karen Hall, Karen Hall, Frederic Rabari-Oskoui, Dan Rubin, and Siegfried Schmidt. In addition, we gratefully acknowledge the biostatistical analyses of Daniel Crusan. Funding was from NIH Pharmacogenetics Research Network grant U01-GM074492; K23 grants HL091120 (A. L. Beitelshees) and HL086558 (R. M. Cooper-DeHoff); CTSA grants UL1-RR092890 (University of Florida), UL1-RR025008 (Emory University), and UL1-RR024150 (Mayo Clinic); and funds from the Mayo Foundation.
refgrp Why Is Out-of-Office Blood Pressure Measurement Needed? Home Blood Pressure Measurements Will Increasingly Replace Ambulatory Blood Pressure Monitoring in the Diagnosis and Management of HypertensionParatiGOmboniSBiloGHypertension200954181lpage 18710.1161/HYPERTENSIONAHA.108.122853Ambulatory blood pressure monitoring and blood pressure self-measurement in the diagnosis and management of hypertensionAppelLJStasonWBAnn Intern Med19931188678828093115Ambulatory blood pressure monitoring and organ damageManciaGParatiGHypertension200036894900link fulltext 11082163Cardiovascular prognosis of "masked hypertension" detected by blood pressure self-measurement in elderly treated hypertensive patientsBobrieGChatellierGGenesNClersonPVaurLVaisseBMenardJMallionJMJAMA20042911342134910.1001/jama.291.11.134215026401Antihypertensive treatment based on blood pressure measurement at home or in the physician's office: a randomized controlled trialStaessenJADen HondECelisHFagardRKearyLVandenhovenGO'BrienETJAMA200429195596410.1001/jama.291.8.95514982911Predicting cardiovascular risk using conventional vs ambulatory blood pressure in older patients with systolic hypertension. Systolic Hypertension in Europe Trial Investigators [see comments]StaessenJAThijsLFagardRO'BrienETClementDde LeeuwPWManciaGNachevCPalatiniPParatiGetal JAMA199928253954610.1001/jama.282.6.53910450715Office compared with ambulatory blood pressure in assessing response to antihypertensive treatment: a meta-analysisManciaGParatiGJ Hypertens20042243544515076144Changes in home versus clinic blood pressure with antihypertensive treatments: a meta-analysisIshikawaJCarrollDJKuruvillaSSchwartzJEPickeringTGHypertension20085285686410.1161/HYPERTENSIONAHA.108.11560018809791Personalized medicine for high blood pressureTurnerSTSchwartzGLBoerwinkleEHypertension2007501510.1161/HYPERTENSIONAHA.107.08704917470720Age-race subgroup compared with renin profile as predictors of blood pressure response to antihypertensive therapy. Department of Veterans Affairs Cooperative Study Group on Antihypertensive AgentsPrestonRAMatersonBJRedaDJWilliamsDWHamburgerRJCushmanWCAndersonRJJAMA19982801168117210.1001/jama.280.13.11689777817Plasma renin activity predicts blood pressure responses to beta-blocker and thiazide diuretic as monotherapy and add-on therapy for hypertensionTurnerSTSchwartzGLChapmanABBeitelsheesALGumsJGCooper-DeHoffRMBoerwinkleEJohnsonJABaileyKRAm J Hypertens2010231014102210.1038/ajh.2010.98pmcid 294169920725057Predictors of antihypertensive response to a standard dose of hydrochlorothiazide for essential hypertensionChapmanABSchwartzGLBoerwinkleETurnerSTKidney Int2002611047105510.1046/j.1523-1755.2002.00200.x11849460Predictors of blood pressure response to the angiotensin receptor blocker candesartan in essential hypertensionCanzanelloVJBaranco-PryorERahbari-OskouiFSchwartzGLBoerwinkleETurnerSTChapmanABAm J Hypertens200821616610.1038/ajh.2007.2418091745Reproducibility of blood pressure response to hydrochlorothiazideFinkielmanJDSchwartzGLChapmanABBoerwinkleETurnerSTJ Clin Hypertens (Greenwich)2002440841210.1111/j.1524-6175.2002.00965.xLack of agreement between office and ambulatory blood pressure responses to hydrochlorothiazideFinkielmanJDSchwartzGLChapmanABBoerwinkleETurnerSTAm J Hypertens20051839840210.1016/j.amjhyper.2004.10.02115797660Comparison of office, ambulatory, and home blood pressure antihypertensive response to atenolol and hydrochlorthiazideBeitelsheesALGongYBaileyKRTurnerSTChapmanABSchwartzGLGumsJGBoerwinkleEJohnsonJAJ Clin Hypertens (Greenwich)201012142110.1111/j.1751-7176.2009.00185.xPharmacogenomics of antihypertensive drugs: rationale and design of the Pharmacogenomic Evaluation of Antihypertensive Responses (PEAR) studyJohnsonJABoerwinkleEZinehIChapmanABBaileyKCooper-DeHoffRMGumsJCurryRWGongYBeitelsheesALAm Heart J200915744244910.1016/j.ahj.2008.11.018267128719249413Validation of two devices for self-measurement of brachial blood pressure according to the International Protocol of the European Society of Hypertension: the SEINEX SE-9400 and the Microlife BP 3AC1-1TopouchianJAEl AssaadMAOrobinskaiaLVEl FeghaliRNAsmarRGBlood Press Monit20051032533110.1097/00126097-200512000-0000816330959Accuracy of the SpaceLabs 90207 determined by the British Hypertension Society protocolO'BrienEMeeFAtkinsNO'MalleyKJ Hypertens1991957357410.1097/00004872-199106000-000161653299Relation between initial blood pressure and its fall with treatmentGillJSZezulkaAVBeeversDGDaviesPLancet198515675692857912SnedecorGWCochranWGStatistical methodspublisher Ames IA,: Iowa State Univ. Pressedition 71980The future of genetic studies of complex human diseasesRischNMerikangasKScience19962731516151710.1126/science.273.5281.15168801636Angiotensin-converting enzyme gene polymorphism and reversibility of uremic left ventricular hypertrophy following long-term antihypertensive therapyCannellaGPaolettiEBarocciSMassarinoFDelfinoRRaveraGDi MaioGNoceraAPatronePRollaDKidney Int19985461862610.1046/j.1523-1755.1998.00027.x9690230Genome-wide association study of blood pressure and hypertensionLevyDEhretGBRiceKVerwoertGCLaunerLJDehghanAGlazerNLMorrisonACJohnsonADAspelundTNat Genet20094167768710.1038/ng.384299871219430479Monitoring initial response to Angiotensin-converting enzyme inhibitor-based regimens: an individual patient data meta-analysis from randomized, placebo-controlled trialsBellKJHayenAMacaskillPCraigJCNealBCFoxKMRemmeWJAsselbergsFWvan GilstWHMacmahonSHypertension20105653353910.1161/HYPERTENSIONAHA.110.15242120625081Next-generation genome-wide association studies: time to focus on phenotype?MacraeCAVasanRSCirc Cardiovasc Genet2011433433610.1161/CIRCGENETICS.111.96076521846867Determinants of blood pressure response to quinapril in black and white hypertensive patients: the Quinapril Titration Interval Management Evaluation trialMokweEOhmitSENasserSAShafiTSaundersECrookEDudleyAFlackJMHypertension2004431202120710.1161/01.HYP.0000127924.67353.8615117912